基于通用学习字典的机织物纹理表征  

Woven Fabric Texture Characterization Based on General Dictionary

作　　者：占竹 王凯[1] 李立轻[1] 汪军[1] ZHAN Zhu;WANG Kai;LI Liqing;WANG Jun(Donghua University,Shanghai,201620,China)

机构地区：[1]东华大学,上海201620

出　　处：《棉纺织技术》2021年第1期34-39,共6页Cotton Textile Technology

基　　金：中央高校基本科研业务费专项资金(CUSF-DF-D-2018039)。

摘　　要：为了提高学习字典的通用性,使其对不同织物密度样本均具有良好重构效果,提出一种包含多种织物纹理特征的通用学习字典构造方法。首先确定不同织物密度样本的取样尺寸和优化用于重构不同织物密度样本的学习字典原子个数,然后依据优化的参数训练得到不同织物密度样本的字典,最后将上述字典进行合并得到通用学习字典。针对不同密度的机织物纹理图像样本分别分析了取样尺寸和字典原子个数对图像重构误差的影响。试验结果表明:通用学习字典对不同机织物纹理的表征性能明显优于离散余弦变换字典。认为:通用学习字典具备了对多种不同密度织物的通用性,重构性好。In order to improve the commonality of dictionary and nice reconstruction effect on different fabric density specimen,a general dictionary construction method was put forward.Firstly,the sample size of fabrics with different density was confirmed and the dictionary atomic number was optimized to reconstruct the samples with different fabric densities.Then,the dictionary with the specimen of different fabric counts were obtained according to optimized parameter training.In the end,general dictionary was achieved by combing the above dictionaries.Aimed at woven texture image specimen with different density,the influence of sample size and dictionary atomic number on image reconstruction error was analyzed respectively.The experiment result showed that the texture characterization of different woven fabrics with dictionary was obviously better than that of discrete cosine transform dictionary.It is considered that general dictionary has the commonality on multiple different densities and the reconstruction property is better.

关 键 词：通用学习字典 纹理表征 计算机视觉 数字图像处理 织物规格 

分 类 号：TS101[轻工技术与工程—纺织工程]